Become the Software Architect you always wanted to be.
Architect up front. Blueprints, components and scope plans tell AI agents what to build, in which order, and what done looks like, so the work comes back the way you designed it.
Everything an agent needs to build it right.
YoStaq is the layer above your agent. You design; it holds the design; the agent builds from it, one step at a time, and reports back.
Blueprints
- Compose the stack on a board: runtime, backend, database, CMS, frontend, capabilities.
- Requires, conflicts and provides are checked as you place them.
- Publish immutable versions. Agents build from a snapshot, never a moving draft.
Skills
- Your conventions as skill files: naming, testing, structure, review.
- Written into the repository, where the agent reads them.
- Fork what is shared, keep yours private.
Components
- Install steps, environment variables and files, per component.
- Your catalog, your conventions: Laravel, Twill, Postgres, Redis, whatever you ship.
- Skills attach to components, so the agent inherits them with the stack.
Scope plans
- Sections tagged to components, then phases and steps.
- Acceptance criteria in EARS form: when X, the system shall Y.
- Coverage shows what the plan still leaves unsaid.
Builds
- The agent takes one step, reports back, takes the next.
- Round trips and corrections per step, an event ledger per build.
- Pause, resume or abandon. The server owns the state.
MCP connection
- One endpoint, four tools. Any agent that speaks MCP can build.
- A token per agent, telemetry per call.
- Builds resume across sessions.
MCP driven. Your agent, your environment.
- YoStaq is not an agent and sells none. Claude Code, Codex or any MCP agent connects to it and does the building.
- The work stays in your editor, your terminal and your repository. YoStaq holds the plan and the state of the build.
- You pay for your agent where you already do. YoStaq is for better planning and a build that stays on track.
An optional assistant, on your own keys.
- The in-app assistant drafts components, sections and criteria, and reviews a blueprint. Use it or leave it closed.
- It runs on your own Anthropic or OpenAI key, stored encrypted, never logged, never shown past the last four characters.
- Skip it and YoStaq adds no AI cost at all.
Straight answers.
What is YoStaq?
The architecture layer above your AI agent. You define blueprints (the stack), components (how each part is installed and configured), skills (your conventions) and scope plans (what to build, with acceptance criteria). Agents connect over MCP and build from that, one step at a time.
What is it not?
Not an IDE, not a code generator and not another agent. YoStaq writes none of your code. It tells the agent you already use what to build, in which order, and what done means.
Who is it for?
Agencies and product teams that ship many similar projects: the same stack, a different scope each time. Every build starts from the same architecture and the same conventions, so the tenth project is as sound as the first.
Which agents work with it?
Claude Code and Codex today. Anything that speaks MCP can connect: one endpoint, a token per agent, four tools.
Does my code go to YoStaq?
No. The agent runs in your repository. YoStaq holds the blueprint, the plan and the state of the build, and the agent reports what it did, step by step.
Do I need my own API keys?
For the in-app assistant, yes: an Anthropic or OpenAI key, stored encrypted and never logged. Your agent keeps using its own subscription or keys; YoStaq never sits between it and the model.
When can I use it?
The launch window is October 2026. Leave your email and you get one message when it opens, nothing before and nothing after. Want to start sooner? Email info@yotech.ai to request early access.
Be first in when invites open.
Private preview first, then a public beta. One email when it is your turn.